Sitk loader

sitk_loader(sample, path_imagedir, image_format=None, grayscale=True, resampling=(1.0, 1.0, 1.0), outside_value=0, **kwargs)

SimpleITK Loader for loading of CT/MRI scans in NIfTI (nii) or Metafile (mha) format within the AUCMEDI pipeline.

The SimpleITK Loader is an IO_loader function, which have to be passed to the DataGenerator.

This loader is intended to load only 3D volumes with annotated voxel spacings.

By default, volumes are normalized to voxel spacing 1.0 x 1.0 x 1.0.
You can define a custom voxel spacing for the loader, by passing a tuple of spacings as parameter 'resampling'.

Info

The SimpleITK Loader utilizes SimpleITK for sample loading:
https://simpleitk.readthedocs.io/en/master/IO.html

Example

# Import required libraries
from aucmedi import *
from aucmedi.data_processing.io_loader import sitk_loader

# Initialize input data reader
ds = input_interface(interface="csv",
                    path_imagedir="dataset/nii_files/",
                    path_data="dataset/annotations.csv",
                    ohe=False, col_sample="ID", col_class="diagnosis")
(samples, class_ohe, nclasses, class_names, image_format) = ds

# Initialize DataGenerator with sitk_loader
data_gen = DataGenerator(samples, "dataset/nii_files/", labels=class_ohe,
                        image_format=image_format, resize=None,
                        grayscale=True, resampling=(2.10, 1.48, 1.48),
                        loader=sitk_loader)

Parameters:

Name	Type	Description	Default
sample	str	Sample name/index of an image.	required
path_imagedir	str	Path to the directory containing the images.	required
image_format	str	Image format to add at the end of the sample index for image loading.	None
grayscale	bool	Boolean, whether images are grayscale or RGB.	True
resampling	tuple of float	Tuple of 3x floats with z,y,x mapping encoding voxel spacing. If passing None, no normalization will be performed.	(1.0, 1.0, 1.0)
**kwargs	dict	Additional parameters for the sample loader.	{}

Source code in aucmedi/data_processing/io_loader/sitk_loader.py

def sitk_loader(sample, path_imagedir, image_format=None, grayscale=True,
                resampling=(1.0, 1.0, 1.0), outside_value=0, **kwargs):
    """ SimpleITK Loader for loading of CT/MRI scans in NIfTI (nii) or Metafile (mha) format within the AUCMEDI pipeline.

    The SimpleITK Loader is an IO_loader function, which have to be passed to the
    [DataGenerator][aucmedi.data_processing.data_generator.DataGenerator].

    This loader is intended to load only 3D volumes with annotated voxel spacings.

    By default, volumes are normalized to voxel spacing 1.0 x 1.0 x 1.0. <br>
    You can define a custom voxel spacing for the loader, by passing a tuple of spacings as parameter 'resampling'. <br>

    ???+ info
        The SimpleITK Loader utilizes SimpleITK for sample loading: <br>
        https://simpleitk.readthedocs.io/en/master/IO.html

    ???+ example
        ```python
        # Import required libraries
        from aucmedi import *
        from aucmedi.data_processing.io_loader import sitk_loader

        # Initialize input data reader
        ds = input_interface(interface="csv",
                            path_imagedir="dataset/nii_files/",
                            path_data="dataset/annotations.csv",
                            ohe=False, col_sample="ID", col_class="diagnosis")
        (samples, class_ohe, nclasses, class_names, image_format) = ds

        # Initialize DataGenerator with sitk_loader
        data_gen = DataGenerator(samples, "dataset/nii_files/", labels=class_ohe,
                                image_format=image_format, resize=None,
                                grayscale=True, resampling=(2.10, 1.48, 1.48),
                                loader=sitk_loader)
        ```

    Args:
        sample (str):               Sample name/index of an image.
        path_imagedir (str):        Path to the directory containing the images.
        image_format (str):         Image format to add at the end of the sample index for image loading.
        grayscale (bool):           Boolean, whether images are grayscale or RGB.
        resampling (tuple of float):Tuple of 3x floats with z,y,x mapping encoding voxel spacing.
                                    If passing `None`, no normalization will be performed.
        **kwargs (dict):            Additional parameters for the sample loader.
    """
    # Get image path
    if image_format : img_file = sample + "." + image_format
    else : img_file = sample
    path_img = os.path.join(path_imagedir, img_file)
    # Load image via the SimpleITK package
    sample_itk = sitk.ReadImage(path_img)
    # Perform resampling
    if resampling is not None:
        # Extract information from sample
        shape = sample_itk.GetSize()
        spacing = sample_itk.GetSpacing()
        # Reverse resampling spacing to sITK mapping (z,y,x -> x,y,z)
        new_spacing = resampling[::-1]
        # Estimate output shape after resampling
        output_shape = []
        for t in zip(shape, spacing, new_spacing):
            s = int(t[0] * t[1] / t[2])
            output_shape.append(s)
        output_shape = tuple(output_shape)
        # Perform resampling via sITK
        sample_itk_resampled = sitk.Resample(sample_itk,
                                             output_shape,
                                             sitk.Transform(),
                                             sitk.sitkLinear,
                                             sample_itk.GetOrigin(),
                                             new_spacing,
                                             sample_itk.GetDirection(),
                                             outside_value)
    # Skip resampling if None
    else : sample_itk_resampled = sample_itk
    # Convert to NumPy
    img = sitk.GetArrayFromImage(sample_itk_resampled)
    # Add single channel axis
    if len(img.shape) == 3 : img = np.expand_dims(img, axis=-1)
    # Return image
    return img